Genes and Immunity (1999) 1, 120–129 1999 Stockton Press All rights reserved 1466-4879/99 $15.00 http://www.stockton-press.co.uk Taxonomic hierarchy of HLA class I allele sequences LM McKenzie1, J Pecon-Slattery1, M Carrington2 and SJ O’Brien1 1Laboratory of Genomic Diversity, Frederick Cancer Research and Development Center, National Cancer Institute, Frederick, MD, 21702, USA; 2Intramural Research Support Program, SAIC-Frederick, National Cancer Institute-Frederick Cancer Research and Development Center, Frederick, MD, 21702, USA The markedly high levels of polymorphism present in classical class I loci of the human major histocompatibility complex have been implicated in infectious and immune disease recognition. The large numbers of alleles present at these loci have, however, limited efforts to verify associations between individual alleles and specific diseases. As an approach to reduce allele diversity to hierarchical evolutionarily related groups, we performed phylogenetic analyses of available HLA- A, B and C allele complete sequences (n = 216 alleles) using different approaches (maximum parsimony, distance-based minimum evolution and maximum likelihood). Full nucleotide and amino acid sequences were considered as well as abridged sequences from the hypervariable peptide binding region, known to interact in vivo, with HLA presented foreign peptide. The consensus analyses revealed robust clusters of 36 HLA-C alleles concordant for full and PBR sequence analyses. HLA-A alleles (n = 60) assorted into 12 groups based on full nucleotide and amino acid sequence which with few exceptions recapitulated serological groupings, however the patterns were largely discordant with clusters prescribed by PBR sequences. HLA-B which has the most alleles (n = 120) and which unlike HLA-A and -C is thought to be subject to frequent recombinational exchange, showed limited phylogenetic structure consistent with recent selection driven retention of maximum heterozygosity and population diversity. Those allele categories recognized offer an explicit phylogenetic criterion for grouping alleles potentially relevant for epidemiologic associations, for inferring the origin of MHC genome organization, and for comparing functional constraints in peptide presentation of HLA alleles. Keywords: evolution; HLA; phylogenetic; peptide binding region Introduction critical for peptide binding and are therefore thought to be critical for T cell recognition9 and represent the func- The human major histocompatibility complex (MHC) is tional part of the molecule. Further, patterns of a multi-gene family comprised of several loci which are mutational substitutions within the PBR have affirmed 1,2 marked by unusually high levels of polymorphism. the strong role of historic selection pressures in preserv- Class I genes (HLA-A, -B and -C) encode molecules which ing functional variation, likely with peptide binding present antigenic peptides on the cell surface for recog- consequences, for this region.8,11–13 + nition by CD8 T cells. Previous research with class I loci To date there are some 108 HLA-A, 223 HLA-B and 67 indicate a correlation between specific alleles and the HLA-C alleles which have been classified into groups or 3–6 genetic basis of resistance to particular diseases. How- families on the basis of serological reactivity.14 For ever, the relationships uniting each of the alleles at these example, HLA-A alleles separate into six serological fam- loci are less clear. Competing processes of selection, ilies; A1/3/11, A9, A2/28, A10, A1915 and A80.16,17 A recombination, functional constraints, and mutation rates small number of studies describe some possible mole- obscure efforts to identify groups of alleles involved with cular relationships among HLA-A, -B and -C alleles 2,7,8 specific diseases. respectively. A preliminary analysis of available cDNA The majority of polymorphic sites in class I alleles are sequences of alleles of the HLA- Aw19 complex (A29, found in exons 2 and 3 which code for the alpha 1 and A30, A31, A32 and A33) reveal that, with the exception alpha 2 domains of the mature protein. These domains of A30 (which is genetically related to the A1/A3/A11 9,10 form the peptide binding region (PBR). Fifty-seven group), seven nucleotides are specific to this group.15 residues within the PBR have been identified as being Further, these Aw19 specific positions were spread over all exons except for the highly variable exon 2. Similarly, an analysis of the nucleotide sequence relationships among 16 HLA-A alleles using maximum parsimony sug- Correspondence: Stephen J O’Brien, PhD, Laboratory of Genomic Diver- gest these alleles cluster according to serological typing, sity, Frederick Cancer Research and Development Center, Building 560, again with the exception of A30 which grouped with the Frederick, MD, 21702, USA. E-mail: [email protected] A3/11/1 family.18 However, this study is limited by The publisher or recipient acknowledges the right of the US assessment of a single allele from each serological group- Government to retain a non-exclusive, royalty-free license in and ing and a lack of statistical analysis of the phylogenetic to any copyright covering the article. This project has been funded in part with Federal funds from the clusters. Likewise, an analysis of complete cDNA National Cancer Institute, National Institutes of Health, under Con- sequences of 14 HLA-C alleles revealed that HLA-C alleles tract No. NO1-CO-56000. divide into two groups, one comprising the Cw7 sero- Taxonomic hierarchy of HLA class I allele sequences LM McKenzie et al 121 type, and the other group including all remaining bootstrap values (as an indication of confidence for the alleles.19 These and other studies have sought to elucidate divergent nodes) generated by the ME and MP analyses the sequence relationships of a subset of HLA class I and significant associations by the ML method. Only ME alleles by parsimony or distance-based phylogenetic trees of nucleotide data are presented as figures in this methods.20,21 manuscript. ME trees of amino acids and MP and ML The extensive polymorphism in HLA class I genes is trees can be viewed on the Laboratory of Genomic Diver- thought to increase the repertoire of immune responses sity (LGD) web site which is located at which an individual can mount. Generation and mainte- http://Igd.nci.nih.gov nance of polymorphism at these loci appear to be the result of selection, interallelic recombination or gene con- HLA-A version.22–24 Present allelic variation has been studied Complete sequence analyses: All five methods of phylogen- extensively to reveal possible associations between dis- etic analysis using complete nucleotide and amino acid ease susceptibility/resistance or disease progression and sequences of HLA-A alleles exhibited a consistent pattern specific HLA alleles. HLA associations have been ident- among trees (Figure 1a, Table 1, LGD web site). Phylo- ified for several infectious diseases including malaria and genetic analyses resulted in a basal bifurcation. Of the HLA-B53,3,4 chronic arthritis and HLA-B276 and various two major evolutionary groups formed, the first includes HLA alleles and the rate of disease progression in HIV.5 alleles of serological families A1, A36, A11, A3, A9, A- .(The large number of class I alleles is responsible, at 8001 and two of ten A19 alleles (herein designated A19ٙ least in part, for low levels of statistical confidence for Specific alleles included in each serological family are reported associations plus the failure to reproduce epide- indicated in Figure 1a after Bodmer et al.14 Within this miological efforts in parallel studies. Here we describe a HLA-A lineage, additional resolution occurred, highly series of phylogenetic analyses of all complete HLA class supported by bootstrap values, in accordance with recog- I allele sequences to identify alleles with a common nized serological terminology. Thus, all five complete evolutionary history. These analyses use patterns of sequence topologies depicted the A9 family as a separate mutations to establish a hierarchical system of grouping subgroup as well as a consistent close association alleles. Whilst any groups identified may reflect evolu- between A1 and A36 allele families (Table 1). Three of tionary relationships among alleles, they may not neces- five analyses indicated an additional association between sarily reflect functional relationships. Hughes et al25 sug- A1, A11 and A36 families. The single allele of the A80 gest that there is a relationship between the PBR region family (A-8001) was placed within this evolutionary clus- of HLA molecules and their role in disease processes. ter in all analyses except the distance-based phylogeny Since the majority of genetic polymorphism between of amino acids. class I alleles is found in the 57 functional residues of The second major evolutionary lineage within HLA-A the PBR, a region implicated in antigen presentation and consisted of families A2, A28, A10, A4301, and the disease association, it was of interest to compare phylo- remainder of the A19 family identified as A19Љ (Figure genetic relationships among the functional residues of the 1a). Although the relative branching order uniting these PBR as well as the complete sequences. distinct families varied with the five methods employed, Here we report a comprehensive phylogenetic analysis the integrity of the family groupings was well-supported of currently available HLA-A, -B and -C alleles using by high bootstrap values (Table 1). The